Hello.
Problem statement: Fedora 17. Updating to recent kernel on Pandaboard ES causes HDMI to stop working. The known good setup is the stock panda-xfce image[1] on the fedora-arm wiki [2]. There are some kernel parameters to mess with the video output, HDMI or DVI-D at omapedia [3], but those do not help. Here is what I tried in /boot/uboot/uEnv.txt: omapfb.mode=dvi:1280x1024MR-24@60 omapdss.def_disp=dvi
Steps to reproduce: yum --nogpgcheck install http://dmarlin.fedorapeople.org/packages/FedoraArm/RPMS/noarch/xpfa-17-1.1.n... yum update grubby yum update kernel-omap ...
Work-around: go back to the old kernel.
cd /boot/uboot cp uImage-3.4.2-3.fc17.armv7hl.omap uImage-omap cp uInitrd-3.4.2-3.fc17.armv7hl.omap uInitrd-omap [reboot]
[1] http://download.fedoraproject.org/pub/fedora-secondary/releases/17/Images/ar... [2] http://fedoraproject.org/wiki/Architectures/ARM/Pandaboard [3] http://omappedia.org/wiki/Bootargs_for_enabling_display#DSS_specific_bootarg...
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
El Mon, 27 Aug 2012 13:15:43 -0500 Jon jdisnard@gmail.com escribió:
Hello.
Problem statement: Fedora 17. Updating to recent kernel on Pandaboard ES causes HDMI to stop working. The known good setup is the stock panda-xfce image[1] on the fedora-arm wiki [2]. There are some kernel parameters to mess with the video output, HDMI or DVI-D at omapedia [3], but those do not help. Here is what I tried in /boot/uboot/uEnv.txt: omapfb.mode=dvi:1280x1024MR-24@60 omapdss.def_disp=dvi
we use omapdrm not omapfb so none of those options will have any effect. its all auto detected and auto setup. if the edid data is not read from the monitor you will get 1024x768 only. ive not tried the 3.5 based kernels as im not at home where i have access to a monitor.
you shouldn't use the xpfa repo there is no need Fedora 17 will update and work just fine with whats in Fedora 17. you could try switching which port you plug the monitor into. you would also need to look at the output of dmesg to see what is happening.
Dennis
Here is what I tried in /boot/uboot/uEnv.txt: omapfb.mode=dvi:1280x1024MR-24@60 omapdss.def_disp=dvi
we use omapdrm not omapfb so none of those options will have any effect. its all auto detected and auto setup. if the edid data is not read from the monitor you will get 1024x768 only. ive not tried the 3.5 based kernels as im not at home where i have access to a monitor.
Thanks Dennis.
The information about omapdrm was good, gave something to troubleshoot.
On the working (3.4.2-3) kernel, you can do this:
# cd /sys/devices/platform/omapdrm.0/drm/card0/card0-HDMI-A-1/
# cat modes 1024x768 800x600 848x480 640x480
On the latest kernel 3.5.2-3 that sysfs path does not exist. :( So I guess the omapdrm driver is not happy.
The output of lsmod does not have as much stuff as the 3.4.2-3 kernel had.
Here I tried to load omapdrm module:
# insmod /lib/modules/3.5.2-3.fc17.armv7hl.omap/kernel/drivers/staging/omapdrm/omapdrm.ko Error: could not insert module /lib/modules/3.5.2-3.fc17.armv7hl.omap/kernel/drivers/staging/omapdrm/omapdrm.ko: Unknown symbol in module
On the working kernel, lsmod: # lsmod Module Size Used by lockd 76647 0 sunrpc 229730 1 lockd aes_generic 31238 0 bnep 10592 2 bluetooth 177728 7 bnep arc4 1269 0 nls_ascii 3508 1 vfat 9199 1 fat 48924 1 vfat wl12xx 145546 0 mac80211 442857 1 wl12xx cfg80211 169876 2 mac80211,wl12xx rfkill 16628 5 cfg80211,bluetooth smsc95xx 12138 0 evbug 1926 0 rtc_twl 5447 0 wl12xx_sdio 3387 0 autofs4 22550 2 omapdrm 62957 2 fb_sys_fops 1376 1 omapdrm syscopyarea 3028 1 omapdrm sysfillrect 3304 1 omapdrm sysimgblt 2131 1 omapdrm drm_kms_helper 28433 1 omapdrm drm 201384 2 drm_kms_helper,omapdrm
And the unhappy kernel... # lsmod Module Size Used by lockd 76997 0 sunrpc 230092 1 lockd bnep 10600 2 bluetooth 179620 7 bnep rfkill 16628 2 bluetooth nls_ascii 3508 1 vfat 9199 1 fat 49092 1 vfat evbug 1926 0 smsc95xx 12162 0 rtc_twl 5447 0 autofs4 22550 2
As you can see there are some differences. :(
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
El Mon, 27 Aug 2012 20:35:06 -0500 Jon jdisnard@gmail.com escribió:
Here is what I tried in /boot/uboot/uEnv.txt: omapfb.mode=dvi:1280x1024MR-24@60 omapdss.def_disp=dvi
we use omapdrm not omapfb so none of those options will have any effect. its all auto detected and auto setup. if the edid data is not read from the monitor you will get 1024x768 only. ive not tried the 3.5 based kernels as im not at home where i have access to a monitor.
Thanks Dennis.
The information about omapdrm was good, gave something to troubleshoot.
On the working (3.4.2-3) kernel, you can do this:
# cd /sys/devices/platform/omapdrm.0/drm/card0/card0-HDMI-A-1/
# cat modes 1024x768 800x600 848x480 640x480
On the latest kernel 3.5.2-3 that sysfs path does not exist. :( So I guess the omapdrm driver is not happy.
The output of lsmod does not have as much stuff as the 3.4.2-3 kernel had.
Here I tried to load omapdrm module:
# insmod /lib/modules/3.5.2-3.fc17.armv7hl.omap/kernel/drivers/staging/omapdrm/omapdrm.ko Error: could not insert module /lib/modules/3.5.2-3.fc17.armv7hl.omap/kernel/drivers/staging/omapdrm/omapdrm.ko: Unknown symbol in module
you should just do "modprobe omapdrm" not insmod the .ko as it takes care of its deps.
do a lsinitrd on the initramfs file in /boot and make sure that it has the omapdrm module in it.
Dennis
# insmod /lib/modules/3.5.2-3.fc17.armv7hl.omap/kernel/drivers/staging/omapdrm/omapdrm.ko Error: could not insert module /lib/modules/3.5.2-3.fc17.armv7hl.omap/kernel/drivers/staging/omapdrm/omapdrm.ko: Unknown symbol in module
you should just do "modprobe omapdrm" not insmod the .ko as it takes care of its deps.
That worked, the omapdrm module loaded with modprobe. Unfortunately that does not cause the HDMI video to suddenly start working. Also, the path in sysfs was not present after loading the module:
/sys/devices/platform/omapdrm.0
do a lsinitrd on the initramfs file in /boot and make sure that it has the omapdrm module in it.
Yes, the initrd has the omapdrm module.
On Tue, Aug 28, 2012 at 3:55 AM, Jon jdisnard@gmail.com wrote:
# insmod /lib/modules/3.5.2-3.fc17.armv7hl.omap/kernel/drivers/staging/omapdrm/omapdrm.ko Error: could not insert module /lib/modules/3.5.2-3.fc17.armv7hl.omap/kernel/drivers/staging/omapdrm/omapdrm.ko: Unknown symbol in module
you should just do "modprobe omapdrm" not insmod the .ko as it takes care of its deps.
That worked, the omapdrm module loaded with modprobe. Unfortunately that does not cause the HDMI video to suddenly start working. Also, the path in sysfs was not present after loading the module:
/sys/devices/platform/omapdrm.0
Thinking back to the 3.5 dev cycle in rawhide there was quite a bit of churn in the OMAP video stuff to do with the DSS module, it's possible we'll need to tweak the F-17 kernel build. Not sure if I'll get time to do that this week.
I want to bring the config restructure back to F-17 but OMAP at least has some issues with USB on that build so I need to get the time to test all the various HW devices we vaguely support such as Panda, Beagle, Trimslice, Highbank, and VExpress. We know the imx at least vaguely works. For 3.5+ we really need to move to DeviceTree and I wonder if that isn't at least part of the issue with the video on 3.5
Peter
On Tue, Aug 28, 2012 at 1:51 AM, Peter Robinson pbrobinson@gmail.com wrote:
On Tue, Aug 28, 2012 at 3:55 AM, Jon jdisnard@gmail.com wrote:
# insmod
/lib/modules/3.5.2-3.fc17.armv7hl.omap/kernel/drivers/staging/omapdrm/omapdrm.ko
Error: could not insert module
/lib/modules/3.5.2-3.fc17.armv7hl.omap/kernel/drivers/staging/omapdrm/omapdrm.ko:
Unknown symbol in module
you should just do "modprobe omapdrm" not insmod the .ko as it takes care of its deps.
That worked, the omapdrm module loaded with modprobe. Unfortunately that does not cause the HDMI video to suddenly start
working.
Also, the path in sysfs was not present after loading the module:
/sys/devices/platform/omapdrm.0
Thinking back to the 3.5 dev cycle in rawhide there was quite a bit of churn in the OMAP video stuff to do with the DSS module, it's possible we'll need to tweak the F-17 kernel build. Not sure if I'll get time to do that this week.
I want to bring the config restructure back to F-17 but OMAP at least has some issues with USB on that build so I need to get the time to test all the various HW devices we vaguely support such as Panda, Beagle, Trimslice, Highbank, and VExpress. We know the imx at least vaguely works. For 3.5+ we really need to move to DeviceTree and I wonder if that isn't at least part of the issue with the video on 3.5
Peter
Howdy,
Diffed the configs for 3.4.2-3 and 3.5.2-3.
The 3.4 kernel had these kernel config's:
CONFIG_PANEL_DVI=m CONFIG_SND_OMAP_SOC_HDMI=y CONFIG_SND_OMAP_SOC_OMAP4_HDMI=y
Could that contribute to the lack of HDMI/DVI-D on pandaboard on the more recent kernels? Possibly in addition to the omapdrm module not loading during boot?
Thanks -Jon (irc masta)
On Fri, Sep 7, 2012 at 12:08 AM, Jon jdisnard@gmail.com wrote:
On Tue, Aug 28, 2012 at 1:51 AM, Peter Robinson pbrobinson@gmail.com wrote:
On Tue, Aug 28, 2012 at 3:55 AM, Jon jdisnard@gmail.com wrote:
# insmod /lib/modules/3.5.2-3.fc17.armv7hl.omap/kernel/drivers/staging/omapdrm/omapdrm.ko Error: could not insert module /lib/modules/3.5.2-3.fc17.armv7hl.omap/kernel/drivers/staging/omapdrm/omapdrm.ko: Unknown symbol in module
you should just do "modprobe omapdrm" not insmod the .ko as it takes care of its deps.
That worked, the omapdrm module loaded with modprobe. Unfortunately that does not cause the HDMI video to suddenly start working. Also, the path in sysfs was not present after loading the module:
/sys/devices/platform/omapdrm.0
Thinking back to the 3.5 dev cycle in rawhide there was quite a bit of churn in the OMAP video stuff to do with the DSS module, it's possible we'll need to tweak the F-17 kernel build. Not sure if I'll get time to do that this week.
I want to bring the config restructure back to F-17 but OMAP at least has some issues with USB on that build so I need to get the time to test all the various HW devices we vaguely support such as Panda, Beagle, Trimslice, Highbank, and VExpress. We know the imx at least vaguely works. For 3.5+ we really need to move to DeviceTree and I wonder if that isn't at least part of the issue with the video on 3.5
Peter
Howdy,
Diffed the configs for 3.4.2-3 and 3.5.2-3.
The 3.4 kernel had these kernel config's:
CONFIG_PANEL_DVI=m
That would likely affect DVI but not HDMI.
CONFIG_SND_OMAP_SOC_HDMI=y CONFIG_SND_OMAP_SOC_OMAP4_HDMI=y
They're the sound components for digital sound over the HDMI port so won't affect video.
Could that contribute to the lack of HDMI/DVI-D on pandaboard on the more recent kernels? Possibly in addition to the omapdrm module not loading during boot?
Partially, I'm just looking at it now while I think of it.
Peter